PR5 DEPLOYMENT

Deployment: Model-agnostic methods

library(randomForest)
library(dplyr)
library(mltools)
library(data.table)
library(pdp)
library(plotly)

horas <- read.csv('hour.csv')
datos <- read.csv('day.csv')
kc_house <- read.csv('kc_house_data.csv')

1.- One dimensional Partial Dependence Plot.

The partial dependence plot shows the marginal effect of a feature on the predicted outcome of a previously fit model.

Apply PDP to the regression example of predicting bike rentals. Fit a random forest approximation for the prediction of bike rentals (cnt). Use the partial dependence plot to visualize the relationships the model learned. Use the slides shown in class as model.

#Siguiendo los pasos de la práctica 3, extraemos lo necesario para el análisis
paso1 <- datos %>% select(workingday,holiday)
datos$season <- as.factor(datos$season)

paso2 <- one_hot(as.data.table(datos$season))
paso2 <- paso2[,-4]
names(paso2) = c('springer','summer','fall')

misty <- datos %>% mutate(misty = case_when(weathersit == 2~ 1,
                                            weathersit != 2 ~ 0))

misty <- misty$misty

rain <- datos %>% mutate(rain = case_when(weathersit == 3| weathersit == 4~ 1,
                                            TRUE ~ 0))

rain <- rain$rain

paso5 <- datos %>% mutate(de_temp = temp*41, de_hum = hum*100, de_windspeed = windspeed*67)
paso5 <- paso5[,c(17,18,19)]

days_since_2011 <- as.numeric(difftime(as.Date(datos$dteday, format = '%Y-%m-%d'),as.Date('2011-01-01',format = '%Y-%m-%d'), units = "days"))

cnt <- datos$cnt

df <- cbind(paso1,paso2, misty,rain,paso5,days_since_2011,cnt)

model_rf <- randomForest(cnt~.,data=df)

pdp_days_since <- pdp::partial(model_rf, pred.var = 'days_since_2011', plot = F)
pdp_temp <- pdp::partial(model_rf, pred.var = 'de_temp', plot = F)
pdp_hum <- pdp::partial(model_rf, pred.var = 'de_hum', plot = F)
pdp_wind <- pdp::partial(model_rf, pred.var = 'de_windspeed', plot = F)

p1 <- ggplot(pdp_days_since, aes(x = days_since_2011, y = yhat)) +
  geom_line() +
  xlab("Days since 2011") + ylab('Partial Dependence')

p2 <- ggplot(pdp_temp, aes(x = de_temp, y = yhat)) +
  geom_line() +
  xlab("Temperature")

p3 <- ggplot(pdp_hum, aes(x = de_hum, y = yhat)) +
  geom_line() +
  xlab("Humidity")

p4 <- ggplot(pdp_wind, aes(x = de_windspeed, y = yhat)) +
  geom_line() +
  xlab("Windspeed")

subplot(p1,p2,p3,p4, shareY = T) %>%  
  layout(annotations = list( 
    list(x = 0.01 , y = 1.07, text = "Days since 2011", showarrow = F, xref='paper', yref='paper'), 
    list(x = 0.3 , y = 1.07, text = "Temperature", showarrow = F, xref='paper', yref='paper'),
    list(x = 0.63 , y = 1.07, text = "Humidity", showarrow = F, xref='paper', yref='paper'),
    list(x = 0.95 , y = 1.07, text = "Windspeed", showarrow = F, xref='paper', yref='paper')))

Analyse the influence of days since 2011, temperature, humidity and wind speed on the predicted bike counts.

A medida que avanza el tiempo, el modelo predice que aumenta el número de bicicletas alquiladas, esto es normal, ya que el servicio se da más a conocer a lo largo del tiempo. For warm but not too hot climates, a large number of rented bikes is predicted. Yet, from temperatures over 27 ºC, the number of rented bikes decrease(too much heat). It appears that cyclists are increasingly inhibited from renting a bike when humidity exceeds 60%. Finally, the windier it gets, the less people like to ride a bike. which is logic. It appears that the model predicts the same from 25 km/h, maybe because there is little training data in that range.